Data Analysis Australia has a long history of working with clients from across the mining, oil and gas sectors to help them gain new insights from their existing data resources and to design and inform effective data collection strategies. As consultant statisticians with experience across many industries and sectors we utilise not only statistical techniques that have proven applications in mining but also adapt techniques traditionally used in other sectors to provide innovative solutions to mining challenges.
Techniques used by Data Analysis Australia’s statisticians for mining applications include:
Regression modelling, including generalised linear models and semi-parametric regression;
Predictive modelling and forecasting;
Model selection algorithms and cross-validation;
Geostatistics and spatial analysis;
Simulation and optimisation;
Cluster analysis to understand groupings amongst ore samples;
Classification and Regression Trees to model key variables by partitioning a data set; and
Experimental design for mineral grade, recovery and hardness testwork programs.
A few examples of our experience in providing solutions and insights to clients from the mining sector are provided below.
Modelling Mineral Recovery Processes
To assist with resource evaluation studies, Data Analysis Australia has developed many models of mineral recovery rates. The models are constructed using linear regression techniques with combinations of chemical, mineralogical and other geological measurements used as variables in the models. The model selection process is driven by both the use of search algorithms and expert statistical judgement. The resulting models are used to understand the recovery level that can be expected from a mineral resource and to compute performance benchmarks for processing circuits.
Calibration of Assay Equipment
Accurate calibration of chemical assay equipment is vital to ensuring the collection of high quality data. Data Analysis Australia developed a series of calibration equations for an X-ray refraction device used to assay several different chemicals simultaneously from a small dataset of test samples.
Data Analysis Australia developed an experimental design of ore blends for a nickel mine in Western Australia. This experimental design considered many factors including the components of the ore blends and the proportions of the various ore types. A “genetic” search algorithm was used to select which combinations of samples should be tested out of the millions of possible combinations and the resulting design allowed the many effects associated with processing blended ores to be quantified whilst minimising the time and resources required for the testwork.
Modelling Rainfall and Water Management
Managing water inventories is a critical task for any mineral processing operation. Data Analysis Australia developed rainfall models for a number of processing sites and incorporated these into simulation tools that predicted the likely range of weather, from extremely low rainfall years to extremely high rainfall years. This in turn enabled informed planning decisions to be made to prevent shortfalls or overflows.
Employee Health Data Collection and Analysis
Data Analysis Australia conducted an online (for office-based staff) and paper-based (for non-office-based staff) employee health survey for a large mining company. The analysis was designed to provide information about the priority areas for health promotion activities across the company. In addition, gaining a better understanding of the obstacles to achieving a healthy lifestyle, and the relationship between different factors such as behaviours, attitudes and health scores is important in the planning of effective health promotion strategies for various target groups.